Merge branch 'release_branch_1_3_0' into 'master'

Release branch 1 3 0

See merge request 3d/PandoraBox/pandora_plugins/plugin_mc-cnn!42

CHANGELOG.md

@@ -2,9 +2,21 @@
 
 ## Unreleased
 
+## 1.3.0a1 (November 2023)
+
+### Added
+
+ - Add function to get band values, the numpy array. [#31]
+
+### Changed
+
+- New format for disparity in the user configuration file. [#33]
+- Update with new API of Pandora compute_cost_volume function. [#38]
+
 ### Fixed
 
  - Deletion of the pip install codecov of githubAction CI.
+ - Remove right_disp_map. [#37]
 
 ## 1.2.0 (March 2023)
 

pandora_plugin_mc_cnn/plugin_mc_cnn.py

@@ -61,6 +61,7 @@ def __init__(self, **cfg: Union[int, str]):
         self._window_size = self.cfg["window_size"]
         self._subpix = self.cfg["subpix"]
         self._band = self.cfg["band"]
+        self._step_col = int(self.cfg["step"])
 
     def check_config(self, **cfg: Union[int, str]) -> Dict[str, Union[int, str]]:
         """
@@ -80,13 +81,17 @@ def check_config(self, **cfg: Union[int, str]) -> Dict[str, Union[int, str]]:
             cfg["model_path"] = self._MODEL_PATH
         if "band" not in cfg:
             cfg["band"] = self._BAND
+        if "step" not in cfg:
+            cfg["step"] = self._STEP_COL  # type: ignore
+
 
         schema = {
             "matching_cost_method": And(str, lambda x: x == "mc_cnn"),
             "window_size": And(int, lambda x: x == 11),
             "subpix": And(int, lambda x: x == 1),
             "model_path": And(str, lambda x: os.path.exists(x)),
             "band": Or(str, lambda input: input is None),
+            "step": And(int, lambda y: y >= 1),
         }
 
         checker = Checker(schema)
@@ -101,31 +106,40 @@ def desc(self):
         print("MC-CNN similarity measure")
 
     def compute_cost_volume(
-        self, img_left: xr.Dataset, img_right: xr.Dataset, disp_min: int, disp_max: int
+        self, img_left: xr.Dataset, img_right: xr.Dataset, grid_disp_min: np.ndarray, grid_disp_max: np.ndarray
     ) -> xr.Dataset:
         """
         Computes the cost volume for a pair of images
 
         :param img_left: left Dataset image
         :type img_left:
             xarray.Dataset containing :
-                - im : 2D (row, col) xarray.DataArray
-                - msk : 2D (row, col) xarray.DataArray
+                - im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32
+                - disparity (optional): 3D (disp, row, col) xarray.DataArray float32
+                - msk (optional): 2D (row, col) xarray.DataArray int16
+                - classif (optional): 3D (band_classif, row, col) xarray.DataArray int16
+                - segm (optional): 2D (row, col) xarray.DataArray int16
         :param img_right: right Dataset image
         :type img_right:
             xarray.Dataset containing :
-                - im : 2D (row, col) xarray.DataArray
-                - msk : 2D (row, col) xarray.DataArray
+                - im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32
+                - disparity (optional): 3D (disp, row, col) xarray.DataArray float32
+                - msk (optional): 2D (row, col) xarray.DataArray int16
+                - classif (optional): 3D (band_classif, row, col) xarray.DataArray int16
+                - segm (optional): 2D (row, col) xarray.DataArray int16
         :param disp_min: minimum disparity
-        :type disp_min: int
+        :type disp_min: np.ndarray
         :param disp_max: maximum disparity
-        :type disp_max: int
+        :type disp_max: np.ndarray
         :return: the cost volume dataset
         :rtype:
             xarray.Dataset, with the data variables:
                 - cost_volume 3D xarray.DataArray (row, col, disp)
                 - confidence_measure 3D xarray.DataArray (row, col, indicator)
         """
+        # Obtain absolute min and max disparities
+        disp_min, disp_max = self.get_min_max_from_grid(grid_disp_min, grid_disp_max)
+
         # check band parameter
         self.check_band_input_mc(img_left, img_right)
 

setup.cfg

@@ -41,8 +41,8 @@ setup_requires =
 # Pandora plugin MC-CNN packages dependencies
 install_requires =
     numpy
-    MCCNN==1.2.0
-    pandora>=1.4.0
+    MCCNN==1.2.1
+    pandora==1.6.0a2
     xarray
     json-checker
 

tests/test_cfg_mccnn_fast.json

@@ -2,13 +2,9 @@
   "input" : {
     "img_left" : "tests/image/left_img.png",
     "img_right" : "tests/image/right_img.png",
-    "disp_min" : -60,
-    "disp_max" : 0
+    "disp_left" : [-60, 0]
   },
   "pipeline": {
-    "right_disp_map": {
-	    "method": "accurate"
-    },
     "matching_cost" : {
         "matching_cost_method": "mc_cnn",
         "subpix": 1,
@@ -30,7 +26,7 @@
        "filter_size": 3
     },
     "validation" : {
-        "validation_method": "cross_checking",
+        "validation_method": "cross_checking_accurate",
         "cross_checking_threshold": 1
     }
   }

tests/test_cfg_mccnn_fast_default_values.json

@@ -2,13 +2,9 @@
   "input" : {
     "img_left" : "tests/image/left_img.png",
     "img_right" : "tests/image/right_img.png",
-    "disp_min" : -60,
-    "disp_max" : 0
+    "disp_left" : [-60, 0]
   },
   "pipeline": {
-    "right_disp_map": {
-	    "method": "accurate"
-    },
     "matching_cost" : {
         "matching_cost_method": "mc_cnn"
     },
@@ -27,7 +23,7 @@
        "filter_size": 3
     },
     "validation" : {
-        "validation_method": "cross_checking",
+        "validation_method": "cross_checking_accurate",
         "cross_checking_threshold": 1
     }
   }

tests/test_cfg_mccnn_fast_multiband.json

@@ -2,13 +2,9 @@
   "input" : {
     "img_left" : "tests/image/left_rgb.tif",
     "img_right" : "tests/image/right_rgb.tif",
-    "disp_min" : -60,
-    "disp_max" : 0
+    "disp_left" : [-60, 0]
   },
   "pipeline": {
-    "right_disp_map": {
-	    "method": "accurate"
-    },
     "matching_cost" : {
         "matching_cost_method": "mc_cnn",
         "subpix": 1,
@@ -31,7 +27,7 @@
        "filter_size": 3
     },
     "validation" : {
-        "validation_method": "cross_checking",
+        "validation_method": "cross_checking_accurate",
         "cross_checking_threshold": 1
     }
   }

tests/test_plugin.py

@@ -34,6 +34,7 @@
 import pandora
 from pandora import matching_cost
 from pandora_plugin_mc_cnn.plugin_mc_cnn import MCCNN, get_band_values  # pylint: disable=unused-import
+from pandora.img_tools import add_disparity
 
 
 # pylint: disable=unsubscriptable-object
@@ -177,6 +178,7 @@ def test_invalidates_cost(self):
         left.attrs["no_data_mask"] = 1
         left.attrs["crs"] = None
         left.attrs["transform"] = None
+        left.pipe(add_disparity, disparity=[-1, 1], window=None)
 
         data = np.zeros((13, 13), dtype=np.float64)
         data += 0.1
@@ -321,13 +323,20 @@ def test_invalidates_cost(self):
                 "model_path": "tests/weights/mc_cnn_fast_mb_weights.pt",
             }
         )
-        cv = matching_cost_.compute_cost_volume(left, right, disp_min=-1, disp_max=1)
+        cv = matching_cost_.compute_cost_volume(
+            img_left=left,
+            img_right=right,
+            grid_disp_min=left["disparity"].sel(band_disp="min"),
+            grid_disp_max=left["disparity"].sel(band_disp="max")
+        )
 
         # Check if the calculated cost volume is equal to the ground truth (same shape and all elements equals)
         np.testing.assert_allclose(cv["cost_volume"].data, cv_before_invali, rtol=1e-06)
 
         # Masked cost volume with pandora function
-        matching_cost_.cv_masked(left, right, cv, -1, 1)
+        matching_cost_.cv_masked(
+            left, right, cv, left["disparity"].sel(band_disp="min"), left["disparity"].sel(band_disp="max")
+        )
         # Check if the calculated cost volume is equal to the ground truth (same shape and all elements equals)
         np.testing.assert_allclose(cv["cost_volume"].data, cv_ground_truth, rtol=1e-06)
         # ------------ Test the method with a secondary mask ( reference mask contains valid pixels ) ------------
@@ -348,6 +357,7 @@ def test_invalidates_cost(self):
         left.attrs["no_data_mask"] = 1
         left.attrs["crs"] = None
         left.attrs["transform"] = None
+        left.pipe(add_disparity, disparity=[-1, 1], window=None)
 
         data = np.zeros((13, 13), dtype=np.float64)
         data += 0.1
@@ -511,13 +521,20 @@ def test_invalidates_cost(self):
                 "model_path": "tests/weights/mc_cnn_fast_mb_weights.pt",
             }
         )
-        cv = matching_cost_.compute_cost_volume(left, right, disp_min=-1, disp_max=1)
+        cv = matching_cost_.compute_cost_volume(
+            img_left=left,
+            img_right=right,
+            grid_disp_min=left["disparity"].sel(band_disp="min"),
+            grid_disp_max=left["disparity"].sel(band_disp="max")
+        )
 
         # Check if the calculated cost volume is equal to the ground truth (same shape and all elements equals)
         np.testing.assert_allclose(cv["cost_volume"].data, cv_before_invali, rtol=1e-06)
 
         # Masked cost volume with pandora function
-        matching_cost_.cv_masked(left, right, cv, -1, 1)
+        matching_cost_.cv_masked(
+            left, right, cv, left["disparity"].sel(band_disp="min"), left["disparity"].sel(band_disp="max")
+        )
 
         # Check if the calculated cost volume is equal to the ground truth (same shape and all elements equals)
         np.testing.assert_allclose(cv["cost_volume"].data, cv_ground_truth, rtol=1e-06)